Every edition starts the same way: a crawl over a curated source list,
supplemented by targeted web search for the gaps the feeds don't cover.
This page is the source list — the publications, repositories, and academic
categories the pipeline watches, and the searches that fill in around them.
For how those raw items become a digest, see the pipeline description on the
About page.
RSS & Atom feeds
Twenty feeds are fetched and parsed every crawl, grouped here by how heavily
each one drives the digest. Titles link to the publication; the pipeline reads
their feeds behind the scenes.
Primary
- Simon Willison's Weblog — daily
posts and curated links with sharp commentary; one of the best independent
voices on LLM tooling and the MCP ecosystem.
- Latent Space — weekly deep dives and
daily AI-news roundups; a leading publication for AI engineering.
- Anthropic News — official
Anthropic announcements, read through a community-maintained RSS feed
(Anthropic publishes no official feed of its own).
- LangChain Changelog — product
and framework news across LangChain, LangGraph, and LangSmith: releases and
major announcements.
- InfoQ AI/ML —
enterprise-grade AI architecture coverage, strong on spec-driven development
and agent protocols.
Valuable
- Ars Technica (AI) — generalist tech
press with an AI desk that consistently produces analysis-grade coverage of
AI policy, model-capability evaluations, and security disclosures.
- Martin Fowler / ThoughtWorks — home of
Birgitta Böckeler's spec-driven-development analysis and the ThoughtWorks
Technology Radar's tracking of agentic patterns.
- Google Developers Blog —
Google's ADK, the A2A protocol, Gemini, and agent-framework releases.
- GitHub Blog — Copilot coding-agent updates,
Spec Kit, and the AGENTS.md standard.
- AWS Architecture Blog —
infrastructure and serverless patterns relevant to running agents in production.
- Hacker News (AI) — high-signal
posts above a points threshold, filtered for keywords like AI agents, context
engineering, MCP, coding agents, and Claude Code; a second, lower-threshold
query widens coverage when the primary feed is unavailable.
Supplementary
- DEV Community (AI Agents) — developer
tutorials and framework comparisons; variable quality, scanned selectively.
- Addy Osmani's Blog — influential writing
on spec-driven development and AI coding best practices.
- InfoQ — QCon coverage — conference
reporting that surfaces enterprise agent case studies.
- Qualys TotalAI Blog — an enterprise
security perspective on AI agent and MCP infrastructure.
- Sonatype Blog — software supply-chain
and dependency security, including attacks on AI/ML infrastructure.
- The Register (AI) —
UK tech press with sharp, skeptical coverage of coding agents and AI security.
- SecurityWeek — security trade press
that covers AI vulnerability discovery and agent security as a dedicated beat.
- Mozilla Blog — the official Mozilla
blog; low-volume, followed for browser-security and AI-vulnerability-discovery
posts.
GitHub repositories & organizations
Eleven repositories are monitored for new releases — and, for some, for
discussion threads and commit activity — alongside four organizations watched
for new projects. Titles link to GitHub.
Autonomous coding agents
- OpenHands —
open-source autonomous agents that resolve GitHub issues inside sandboxed
Docker; the closest open-source analog to a full autonomous development platform.
- Aider — terminal-native
AI pair programming with deep git integration; a reference point for CLI agent UX.
Agent frameworks
- LangGraph — graph-based
agent orchestration with durable state and checkpointing.
- CrewAI — a role-based
multi-agent framework (crews and flows) with A2A support.
- 12 Factor Agents —
principles for production agent engineering.
Protocols & standards
- MCP Specification —
the Model Context Protocol specification itself, now under the Agentic AI Foundation.
- MCP Servers —
official reference implementations of MCP servers.
- A2A Protocol — the
Agent-to-Agent communication protocol, under the Linux Foundation.
- AGENTS.md — the open format
for AI coding-agent instructions, under the Agentic AI Foundation.
Spec-driven development
Context engineering
Organizations watched for new repositories
Academic papers — arXiv
Each crawl harvests the previous seven days of submissions from
arXiv via its OAI-PMH bulk interface, across
seven computer-science categories:
- Artificial Intelligence (cs.AI)
- Computation and Language (cs.CL)
- Software Engineering (cs.SE)
- Multiagent Systems (cs.MA)
- Machine Learning (cs.LG)
- Human-Computer Interaction (cs.HC) — where human-in-the-loop research lives
- Cryptography and Security (cs.CR) — where agent-security work appears
That raw harvest is large, so each paper's title and abstract is matched against
a set of keyword themes; only papers that match are kept, and the match assigns
the paper its tags and priority. The themes are: context engineering for agents,
autonomous coding agents, multi-agent software development, spec-driven
development, LLM agent orchestration, AI agent safety and governance, agent
evaluation and benchmarks, and retrieval-augmented generation for agents.
Curated references
Beyond the live crawl, a small set of landmark articles and reports forms the
background reading the digest returns to when a current development echoes an
earlier idea.
- Context Engineering for AI Agents: Lessons from Building Manus —
KV-cache hit rate as the number-one production metric, a 100:1 input/output
token ratio, and a framework rebuilt four times; hard-won, practical context
engineering.
- 12 Factor Agents —
production agent-engineering principles: own your prompts, context, and control
flow, and favour small, focused agents.
- Understanding SDD: Kiro, spec-kit, and Tessl —
Birgitta Böckeler's candid practitioner review of SDD tools: where agents
ignore specs, where they overreach, and the tension in between.
- State of AI Agents (LangChain) —
a survey of 1,300+ respondents: 57% run agents in production, quality is the
top barrier, 89% have observability.
- AI Coding Landscape 2026 (ToolShelf) —
a landscape tracking 200+ AI coding tools, the shift toward terminal-native
agents, and the dominance of open source.
- GitHub Octoverse — GitHub's annual
report: 4.3M AI-related repositories (up 178% year over year) and ecosystem
sizing data.
- Agentic AI Foundation Announcement —
the Linux Foundation fund that now stewards MCP, AGENTS.md, and goose.
- SDD on InfoQ: Architecture Becomes Executable —
frames spec-driven development as an architectural pattern: drift detection,
continuous validation, AI governance.
- Best AI Coding Agents 2026 (Faros AI) —
developer-focused reviews of Claude Code, Cursor, Codex, Copilot, and Cline
against enterprise criteria.
LLM-assisted web search
Feeds and repositories cover most of the ground, but not all of it — some of the
most important sources don't publish a feed at all. To close those gaps, the
crawl runs a handful of targeted web searches every week:
- Anthropic's engineering and research posts. Major product
launches and safety disclosures often land on pages that have no RSS feed, so
each crawl searches for recent posts directly.
- AI safety and security disclosures. Sandbox escapes,
jailbreaks, and vulnerability-discovery stories tend to break in the security
trade press before they reach mainstream AI coverage, so each crawl looks for
the week's disclosures explicitly.
- Pricing and plan changes. Vendors frequently update pricing
quietly, without an announcement, so the crawl checks when a fetch hints that
something has changed.
Separately, the Scouts — the Grimoire's deep-dive briefings — are research-driven
from the start: each one runs at least five topic-shaped searches and fetches the
full text of the most important sources rather than working from headlines.
Search results are useful but slippery: a result gives a headline and a link, not
the article itself. The pipeline's discipline is that every source a claim draws
on has to be captured as a real, fetched source — never a quote pulled from a URL
or a search snippet. That rule, and the failure modes it guards against, are
described on the About page.
Suggest a source
This list is meant to evolve, and good suggestions have changed it before. If you
know a feed, repository, researcher, or publication that belongs here — something
with genuine signal on harness engineering, context engineering, agent
orchestration, agentic protocols, autonomous coding agents, governance, or
spec-driven development — send it to
grimoire@artificerdigital.com.
The same address is where to report anything that looks wrong in an edition.
— Tim